Touhid User

I’m an AI Engineer and IEEE-published researcher (ICCCNT 2025, IIT Indore) with end-to-end AI/ML experience across the full lifecycle—from data preprocessing and model development to deployment and explainability. I build production-ready systems spanning classical ML (XGBoost, Random Forest, LightGBM) and deep learning (CNN, ANN), and I also develop modern LLM/RAG pipelines using LangChain, FAISS, and the Google Gemini API. I focus on accuracy plus interpretability, using SHAP/LIME to make decisions transparent for real stakeholders.

My research and projects reflect that engineering mindset: I developed a CNN-based multi-class TeaGuard-AI for tea leaf disease classification for IEEE ICCCNT 2025, and I’ve built RAG apps with multi-PDF upload, semantic chunking, vector search, and robust Gemini integration (fallback and rate-limit handling) using Streamlit. For applied ML, I created an explainable financial fraud detection pipeline with SHAP-based decision transparency and reported 98–99% ROC-AUC and 99% accuracy across models, plus churn prediction with SHAP/LIME explainability and an interactive Streamlit dashboard (ChurnIQ) with a best model around 91.25% accuracy. Overall, I combine research rigor with engineering discipline to deliver AI solutions that are accurate, interpretable, and built to scale.